CN111143308A - Federal learning-based high-low voltage motor data processing method, system and device - Google Patents
Federal learning-based high-low voltage motor data processing method, system and device Download PDFInfo
- Publication number
- CN111143308A CN111143308A CN201911365903.9A CN201911365903A CN111143308A CN 111143308 A CN111143308 A CN 111143308A CN 201911365903 A CN201911365903 A CN 201911365903A CN 111143308 A CN111143308 A CN 111143308A
- Authority
- CN
- China
- Prior art keywords
- training
- node device
- computing node
- model
- federal
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/10—File systems; File servers
- G06F16/17—Details of further file system functions
- G06F16/176—Support for shared access to files; File sharing support
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
- G06F16/24—Querying
- G06F16/245—Query processing
- G06F16/2458—Special types of queries, e.g. statistical queries, fuzzy queries or distributed queries
- G06F16/2462—Approximate or statistical queries
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Software Systems (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Computation (AREA)
- Mathematical Physics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Databases & Information Systems (AREA)
- Probability & Statistics with Applications (AREA)
- Bioinformatics & Computational Biology (AREA)
- Medical Informatics (AREA)
- Computing Systems (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Fuzzy Systems (AREA)
- Computational Linguistics (AREA)
- Control Of Electric Motors In General (AREA)
Abstract
The invention relates to a high-low voltage motor data processing method, a system and a device based on federal learning. The method comprises the following steps: obtaining a training sample; after the type of the training model is determined, starting a training algorithm built in a federated computing node device, and extracting characteristic independent training; and encrypting and uploading the training result to a cloud server for summary calculation, homomorphic encrypting and calculating the average gradient and respective loss by the cloud server, returning the average gradient and respective loss to each computing node device in an encryption mode, updating each device model after decrypting, and repeating for many times until the model is stable. And when the running state of the motor is predicted, predicting the corresponding motor fault state of the real-time waveform data transmitted to the device according to the optimal model. According to the method, a federal learning framework is adopted, so that the private data of each enterprise can be well protected, and meanwhile, the accuracy of model training is improved.
Description
Technical Field
The invention relates to the field of application of the Internet of things of motor protectors, in particular to a method, a system and a device for processing high-low voltage motor data based on federal learning.
Background
The motor protector is used for comprehensively protecting and controlling the motor. At present, a motor protector usually collects electric parameter data such as three-phase current, three-phase voltage, electric energy quality and the like, and gives an alarm or cuts off a motor power supply to execute protection action when the motor has the problems of overcurrent, undercurrent, phase failure, locked rotor, short circuit, overvoltage, undervoltage, electric leakage, three-phase imbalance, overheating, grounding, bearing abrasion, stator and rotor eccentricity, winding aging and the like.
A large number of motors exist in industrial enterprises such as manufacturing industry, and along with the upgrading of intelligent industry, the protection of the motors becomes a hard demand of the enterprises. The motor protector often analyzes the running state of the motor by collecting three-phase current and three-phase voltage of various motor devices (including low-voltage motors, various fans, pumps and the like, high-voltage motors). Therefore, in theory, an enterprise can analyze the operating state of the motor by analyzing various data collected by the motor protector. Moreover, enterprises increasingly strengthen the fault protection requirements of the motors, and how to effectively utilize data of different high-low voltage motors and predict and analyze real-time fault states of the motors becomes a problem which needs to be solved urgently.
However, because of the small number of data samples of the same enterprise, the data features that can be extracted are limited, and the motor operation state information analyzed by means of the limited data is unreliable. Therefore, the data training can be performed by summarizing the enterprise data through the cloud server, which however causes two problems, namely, the network bandwidth pressure is large due to the large amount of raw data. Secondly, the business or technical information of the enterprises can be leaked due to the fact that the cloud server aggregates the data of the enterprises. With the increasing protection awareness of national and local enterprises on data privacy, the tendency that data cannot be analyzed locally becomes data analysis in the future industrial field, and therefore, a great obstacle exists in gathering data of each enterprise.
Disclosure of Invention
The invention aims to provide a method, a system and a device for processing high-low voltage motor data based on federal learning.
The scheme of the invention comprises the following steps:
a high-low voltage motor data processing method based on federal learning comprises the following steps:
acquiring a training sample, wherein the training sample comprises motor data and a motor state; training samples are stored in a federal computing node device;
the cloud server sends model training instructions to each enterprise center server, wherein the model training instructions comprise training model types;
each enterprise center server forwards the model training instruction to a federal computing node device in an enterprise;
the federal computing node device starts a built-in training algorithm according to the model training instruction, extracts corresponding original data characteristics from a training sample and trains the original data characteristics; encrypting and uploading the training result to a cloud server;
the cloud server performs summary calculation, and sends a summary calculation result to the federal calculation node device through the enterprise center server;
the federal computing node device updates a training model of the federal computing node device according to the result of the summary computation;
repeating the steps, and after multiple rounds of iteration, enabling the training model effect of each federal computing node device to reach a preset standard to obtain an optimal model;
and the federal computing node device predicts the current motor running state by using the optimal model according to the real-time motor data waveform.
Preferably, the training sample is obtained by calibration.
Preferably, the training sample calibration method includes: an operator initiates a manual inspection command, the federal computing node device records a voltage and current data waveform at the moment when the manual training command is sent out, meanwhile, an effective value of the voltage and current data waveform is transmitted to an enterprise central server for backup, the operator gives a state evaluation result according to the current motor state, and meanwhile, the evaluation result is sent to the federal computing node device through the enterprise central server and forms effective training data with the voltage and current data waveform; that piece of data on the central server is the index data.
Preferably, the original data features include standard deviation, variance, kurtosis, and RMS root mean square value; the types of the original data characteristics of the motors of the same type are the same.
Preferably, the training result comprises a gradient.
Preferably, the result of the summary calculation comprises an average gradient.
Preferably, the model training instruction further comprises an encrypted public key; and the training result is encrypted and uploaded according to the encrypted public key.
Preferably, the predetermined criteria include: the loss value of the evaluation index loss function converges or reaches a preset loss threshold.
A high-low voltage motor data processing system based on federal learning comprises cloud servers and enterprise networks of enterprises, wherein the enterprise networks comprise at least one enterprise center server and a plurality of federal computing node devices;
the method comprises the steps that a federal computing node device obtains a training sample, wherein the training sample comprises motor data and a motor state;
training samples are stored in a federal computing node device;
the cloud server sends model training instructions to each enterprise center server, wherein the model training instructions comprise training model types;
each enterprise center server forwards the model training instruction to a federal computing node device in an enterprise;
the federal computing node device starts a built-in training algorithm according to the model training instruction, extracts corresponding original data characteristics from a training sample and trains the original data characteristics; encrypting and uploading the training result to a cloud server; the cloud server performs summary calculation, and sends a summary calculation result to the federal calculation node device through the enterprise center server;
the federal computing node device updates a training model of the federal computing node device according to the result of the summary computation;
repeating the steps, and after multiple rounds of iteration, enabling the training model effect of each federal computing node device to reach a preset standard to obtain an optimal model;
and the federal computing node device predicts the current motor running state by using the optimal model according to the real-time motor data waveform.
A federated compute node apparatus comprising a processor and a memory, processing executing a computer program to implement a method comprising:
the method comprises the steps that a federal computing node device obtains a training sample, wherein the training sample comprises motor data and a motor state; training samples are stored in a federal computing node device;
the method comprises the following steps that a federal computing node device starts a built-in training algorithm according to a model training instruction sent by a cloud server, extracts corresponding original data characteristics from a training sample and trains the corresponding original data characteristics; encrypting and uploading the training result to a cloud server;
the federal computing node device updates a training model of the federal computing node device according to the result of the summary computing of the cloud server;
repeating the steps, and after multiple rounds of iteration, enabling the training model effect of each federal computing node device to reach a preset standard to obtain an optimal model;
and the federal computing node device predicts the current motor running state by using the optimal model according to the real-time motor data waveform.
According to the method, a federal learning framework is adopted, so that the private data of each enterprise can be well protected, and meanwhile, the accuracy of model training is improved.
Drawings
FIG. 1 is a system schematic of an embodiment of the invention;
FIG. 2 is a diagram of a single plant network configuration of an embodiment of the present invention;
FIG. 3 is a schematic diagram of a federated compute node arrangement of an embodiment of the present invention;
FIG. 4 is a flow chart of the preliminary preparation work and method.
Detailed Description
At present, fault protection requirements of a motor are increasingly strengthened by a plurality of manufacturing enterprises, such as industries of chemical industry, electric power, metallurgy and the like, and how to effectively utilize data of different high-low voltage motors and predict and analyze real-time fault states of the motors becomes a problem which needs to be solved urgently. However, in the face of the data islanding problem, each enterprise is not willing to expose own data, and meanwhile, the efficiency of the own data model is urgently needed to be improved, and the federal learning framework becomes an effective means for solving the problem of the industry. Federal Learning, namely federal machine Learning (fed machine Learning), is a machine Learning framework, and can effectively help a plurality of organizations to carry out data use and machine Learning modeling under the condition of meeting the requirements of user privacy protection, data safety and government regulations.
Based on this, the basic idea of the invention is: each production enterprise has many high-low voltage motors of the same type, and need carry out data acquisition and state analysis to it separately, the sample characteristics that it obtained are similar basically, so can draw reference to horizontal federal study's training thinking and design, because the various motor samples of single enterprise are less, it helps promoting model training efficiency to unite other enterprises to carry out model training, both can effectively solve the problem that training data set quantity and characteristic are not enough in the actual industrial environment, also effectively protected the sensitive data privacy problem of each enterprise simultaneously, for the large-scale landing in the industrial field of actual artificial intelligence algorithm provides probably.
The following detailed description is made with reference to the accompanying drawings.
As shown in fig. 1, the system level includes a cloud server, an enterprise center server, a federal computing node device, a motor protector, a sensor and the like. The cloud server is in communication connection with a plurality of enterprise networks; for example, in fig. 1, the cloud server is communicatively connected to an enterprise network a, an enterprise network B, or even an enterprise network X. Each enterprise network comprises a central server and a plurality of federal computing node devices, each federal computing node device is connected with a plurality of motor protectors, and each motor protector collects various information through a corresponding sensor.
As shown in fig. 2, in order to perform unified management, storage and calculation on the motor protector data at the site, one or more federal computing node devices (such as a CAN manager in fig. 2) customized for the industrial field are installed in each production workshop, and the device has the advantages of electromagnetic interference resistance, shock resistance, impact resistance, water resistance, dust resistance and low power consumption, and each federal computing node CAN manage not less than 20 motor protectors.
As shown in fig. 3, the federal computing node device includes a communication module, a storage module, a computing module, an encryption/decryption module, a multi-core low-power CPU, and a reinforcement module.
The communication module is used for realizing two kinds of communications, and the data communication between the motor protector to the federal calculation node device is used for firstly, and coding decoding etc. CAN carry out data communication through multiple communication methods such as TCP/IP agreement, Modbus agreement, CAN bus, to the place that motor protector and federal calculation node device CAN't carry out wired connection, CAN erect signal amplifier and carry out data communication through the 4G module. And the second mode is used for data communication and command transmission from the federal computing node device to an enterprise central server, and supports various networking modes such as bidirectional Ethernet transmission, 4G or GPRS wireless transmission, ZigBee and the like.
The memory module may flexibly configure the ROM capacity to accommodate the configuration of the federated compute node device.
The computing module can be designed by adopting an independent computing chip, and can also utilize a multi-core low-power-consumption CPU of the device to carry out computation.
And the encryption and decryption module is used for homomorphic encryption of the calculation result and transmitting the result to the cloud server side for gathering.
The multi-core low-power consumption CPU is the core of the device and is used for multiple functions such as data processing and the like.
The reinforcing module is a peripheral hardware design module of the device, and achieves the characteristics of electromagnetic interference resistance, shock resistance, impact resistance, water resistance, dust resistance and the like by means of sealing, isolation, shock absorption and the like.
The enterprise central server is used by the enterprise production management system and comprises monitoring data and states of all production links of an enterprise. In this embodiment, the enterprise center server is mainly used for data calibration and forwarding of interactive data between the cloud server and the federal computing node device.
Different from other industries, the sample data and the state of the high-low voltage motor in the industrial field need to be calibrated manually by an operator. In order to obtain a valid training sample, calibration of the sample data is required. An operator can initiate a manual inspection command through a mobile APP developed by the system, a webpage end or a computer application program, and can inspect any high-low voltage motor. When a command is sent out, the federal computing node device records voltage and current and other data waveforms at the moment of sending the command, meanwhile, effective values of the waveforms are transmitted to an enterprise central server for backup, an operator gives a state evaluation result as a label according to the current motor state, the state evaluation result is returned to the enterprise central server for data binding through a mobile APP, a webpage end or a computer application program, and meanwhile, the evaluation result (label) is also sent to the federal computing node device through the enterprise central server, forms effective training data with the recorded waveforms and stores the effective training data in a built-in database of the federal computing node device; and the data on the central server is index data and is stored in an index database of the central server. For the current and voltage data of a certain motor in a certain time period, operation and maintenance personnel need to mark the motor according to the current state of the motor, such as normal state, bearing fault, short circuit, looseness and the like. Such a valid sample is generated and thus the valid sample is the motor data power-up motor state. This step ensures the validity of the data. The data are trained by a plurality of enterprises together, all the data are combined and shared, the number of training samples is considerable, and the reliability and the accuracy of a training model are guaranteed. Meanwhile, the data privacy protection of federal learning also well solves the data security problem brought by common modeling of enterprises.
As shown in fig. 4, the training process is as follows:
the cloud server management platform uniformly sends model training instructions to each enterprise central server, wherein the model training instructions comprise model types, encrypted public keys and the like, and are forwarded to federal computing node devices in enterprises by each enterprise central server, and each federal computing node device is internally provided with each model training algorithm script; the model training algorithm script includes: at present, mainstream machine learning and deep learning algorithms such as decision trees, support vector machines, neural networks and the like are adopted.
After the federal computing node device obtains the instruction, a built-in training algorithm is started, a sample calibrated before is obtained from a built-in database, and each federal computing node device starts to extract original data characteristics according to model types, such as: the method comprises the steps of standard deviation, variance, kurtosis, RMS root mean square value and the like, the types of original data features of motors of the same type are the same, corresponding model training algorithms are called for training, data of high-voltage and low-voltage motors of different types are trained independently, and the gradient of each model is calculated. Each federal computing node device trains a primary model by using sample data of the federal computing node device to obtain primary gradient and loss.
And each computing node encrypts the training gradients of the different types of motor data according to the encryption public key in the model training instruction of the cloud server and uploads the encrypted training gradients to the cloud server. The encryption method used in this embodiment is a homomorphic encryption method.
After the cloud server receives the encrypted gradient data of the motors of different types, the average gradient of the motors of different types is obtained through calculation, the calculation process is all homomorphic encryption calculation, and specific plaintext data cannot be obtained.
And after the server calculates the average gradient, the average gradient is issued to each enterprise network, and the corresponding federal computing node device is found by indexing the database through each enterprise center server. And the federal computing node device updates the coefficients of the orders of the training model according to the average gradient transmitted by the server to obtain an updated model, calculates according to the updated model by using the sample data of the federal computing node device again to obtain the loss and the gradient of the round, uploads the loss and the gradient to the cloud server again, and then the cloud server calculates the average gradient again and issues the average gradient.
After multiple iterations, the loss is reduced, the model is considered to reach an ideal optimal state under the condition that the loss convergence does not reduce any more or a preset loss threshold is met, a high-order equation which can approximately represent the corresponding relation between the data and the motor state is obtained, and the model training process is finished.
The final model was established and immediately placed into service. The federal computing node device acquires the running waveform data of the motor in real time, does not need to store the running waveform data, and predicts the current running state of the motor through a trained model.
Wherein, the gradient and the loss refer to: the training algorithm essentially constructs a high-order equation, which corresponds the input data and the output state. In each training, data is input into a high-order equation of a model structure, and the obtained output result and the real output result of the sample have difference, namely loss, namely the loss value of the evaluation index loss function. Reducing the loss means that the closer the equation output result and the true result are, the more accurate the predicted state is. Calculating the gradient is to solve the high-order equation, and obtain the optimal value of each order coefficient of the high-order equation through the gradient, so as to obtain the global minimum value of the high-order equation, thereby reducing the loss to the minimum.
When the model reaches the ideal state, the evaluation index loss function of the training algorithm is converged, or reaches a set threshold range, which indicates that the data training has already approached the optimal state or a relative optimal state set manually, and the training process can be terminated. The manual setting of the threshold range is to reduce the complexity of calculation, and when the loss function does not converge and oscillate or the convergence speed is slow, the preset threshold can effectively reduce the number of iterations and reduce the calculation time. The gradient, the evaluation index loss function, the loss and the like belong to inherent parameters of a specific training algorithm, and the corresponding specific forms are different according to different algorithm types. However, these arrangements are well known in the art and can be freely designed by those skilled in the relevant art.
The homomorphic encryption calculation is a data encryption mode, the encrypted data can be directly added and multiplied under the condition of not being decrypted, the obtained result is still the encrypted data, and the decrypted result is the same as the result obtained by operating the original data. Homomorphic cryptographic computations are well known in the art and will not be described herein.
As shown in fig. 4, the training algorithm preparation work includes system hardware construction, system software development, and training sample calibration. Training sample calibration has been described above. The system hardware construction comprises construction and debugging of a sensor, a motor protector, a federal computing node device and an enterprise center server. The system software development comprises a mobile APP client, a webpage end, a computer application program and the like which are matched, and the computer application program comprises functions of data display, fault recording and the like.
According to the above description, the data is trained from the beginning to the end to obtain the optimal model, and the optimal model is not left locally, but the effect of multi-sample combined training is achieved, so that the data privacy of each enterprise is effectively protected, and the model efficiency is improved. Therefore, the invention adopts a federal learning framework, can better protect the private data of each enterprise, eliminates the barrier of gathering the data of each enterprise in the background technology, and greatly improves the accuracy of model training because the samples are rich and the model is continuously corrected through average gradient.
Claims (10)
1. A high-low voltage motor data processing method based on federal learning is characterized in that: the method comprises the following steps:
acquiring a training sample, wherein the training sample comprises motor data and a motor state; training samples are stored in a federal computing node device;
the cloud server sends model training instructions to each enterprise center server, wherein the model training instructions comprise training model types;
each enterprise center server forwards the model training instruction to a federal computing node device in an enterprise;
the federal computing node device starts a built-in training algorithm according to the model training instruction, extracts corresponding original data characteristics from a training sample and trains the original data characteristics; encrypting and uploading the training result to a cloud server;
the cloud server performs summary calculation, and sends a summary calculation result to the federal calculation node device through the enterprise center server;
the federal computing node device updates a training model of the federal computing node device according to the result of the summary computation;
repeating the steps, and after multiple rounds of iteration, enabling the training model effect of each federal computing node device to reach a preset standard to obtain an optimal model;
and the federal computing node device predicts the current motor running state by using the optimal model according to the real-time motor data waveform.
2. The federally-learned high-low voltage motor data processing method as claimed in claim 1, wherein: the training samples are obtained by calibration.
3. The federally-learned high-low voltage motor data processing method as claimed in claim 1, wherein: the training sample calibration method comprises the following steps: an operator initiates a manual inspection command, the federal computing node device records a voltage and current data waveform at the moment when the manual training command is sent out, meanwhile, an effective value of the voltage and current data waveform is transmitted to an enterprise central server for backup, the operator gives a state evaluation result according to the current motor state, and meanwhile, the evaluation result is sent to the federal computing node device through the enterprise central server and forms effective training data with the voltage and current data waveform; that piece of data on the central server is the index data.
4. The federally-learned high-low voltage motor data processing method as claimed in claim 1, wherein: the original data features comprise standard deviation, variance, kurtosis and RMS root mean square value; the types of the original data characteristics of the motors of the same type are the same.
5. The federally-learned high-low voltage motor data processing method as claimed in claim 4, wherein: the training results include a gradient.
6. The federally-learned high-low voltage motor data processing method as claimed in claim 5, wherein: the result of the summary calculation includes an average gradient.
7. The federally-learned high-low voltage motor data processing method as claimed in claim 1, wherein: the model training instruction further comprises an encrypted public key; and the training result is encrypted and uploaded according to the encrypted public key.
8. The federally learned high-low voltage motor data processing method as claimed in claim 6, wherein: the predetermined criteria include: the loss value of the evaluation index loss function converges or reaches a preset loss threshold.
9. A high-low voltage motor data processing system based on federal learning is characterized in that: the enterprise network comprises at least one enterprise center server and a plurality of federal computing node devices;
the method comprises the steps that a federal computing node device obtains a training sample, wherein the training sample comprises motor data and a motor state;
training samples are stored in a federal computing node device;
the cloud server sends model training instructions to each enterprise center server, wherein the model training instructions comprise training model types;
each enterprise center server forwards the model training instruction to a federal computing node device in an enterprise;
the federal computing node device starts a built-in training algorithm according to the model training instruction, extracts corresponding original data characteristics from a training sample and trains the original data characteristics; encrypting and uploading the training result to a cloud server; the cloud server performs summary calculation, and sends a summary calculation result to the federal calculation node device through the enterprise center server;
the federal computing node device updates a training model of the federal computing node device according to the result of the summary computation;
repeating the steps, and after multiple rounds of iteration, enabling the training model effect of each federal computing node device to reach a preset standard to obtain an optimal model;
and the federal computing node device predicts the current motor running state by using the optimal model according to the real-time motor data waveform.
10. A federated compute node device, comprising a processor and a memory, characterized in that: processing executes the computer program to implement the method of:
the method comprises the steps that a federal computing node device obtains a training sample, wherein the training sample comprises motor data and a motor state; training samples are stored in a federal computing node device;
the method comprises the following steps that a federal computing node device starts a built-in training algorithm according to a model training instruction sent by a cloud server, extracts corresponding original data characteristics from a training sample and trains the corresponding original data characteristics; encrypting and uploading the training result to a cloud server;
the federal computing node device updates a training model of the federal computing node device according to the result of the summary computing of the cloud server;
repeating the steps, and after multiple rounds of iteration, enabling the training model effect of each federal computing node device to reach a preset standard to obtain an optimal model;
and the federal computing node device predicts the current motor running state by using the optimal model according to the real-time motor data waveform.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911365903.9A CN111143308B (en) | 2019-12-26 | 2019-12-26 | High-low voltage motor data processing method, system and device based on federal learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911365903.9A CN111143308B (en) | 2019-12-26 | 2019-12-26 | High-low voltage motor data processing method, system and device based on federal learning |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111143308A true CN111143308A (en) | 2020-05-12 |
| CN111143308B CN111143308B (en) | 2023-04-28 |
Family
ID=70520611
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201911365903.9A Active CN111143308B (en) | 2019-12-26 | 2019-12-26 | High-low voltage motor data processing method, system and device based on federal learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111143308B (en) |
Cited By (16)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111678696A (en) * | 2020-06-17 | 2020-09-18 | 南昌航空大学 | Intelligent mechanical fault diagnosis method based on federal learning |
| CN111723942A (en) * | 2020-06-29 | 2020-09-29 | 南方电网科学研究院有限责任公司 | Enterprise power load prediction method, power grid service subsystem and prediction system |
| CN111722043A (en) * | 2020-06-29 | 2020-09-29 | 南方电网科学研究院有限责任公司 | Power equipment fault detection method, device and system |
| CN111737749A (en) * | 2020-06-28 | 2020-10-02 | 南方电网科学研究院有限责任公司 | Measuring device alarm prediction method and device based on federal learning |
| CN111796811A (en) * | 2020-07-08 | 2020-10-20 | 上海富数科技有限公司 | Flow control engine system for realizing support of breakpoint debugging in federal learning |
| CN111856934A (en) * | 2020-07-16 | 2020-10-30 | 南京大量数控科技有限公司 | Federal learning data processing algorithm between isomorphic intelligent workshops |
| CN112085074A (en) * | 2020-08-25 | 2020-12-15 | 腾讯科技（深圳）有限公司 | Model parameter updating system, method and device |
| CN112763845A (en) * | 2020-12-23 | 2021-05-07 | 广东电网有限责任公司梅州供电局 | Edge Internet of things firmware fault detection method and system based on federal learning |
| CN113033605A (en) * | 2021-02-08 | 2021-06-25 | 广东迅科动力科技有限公司 | Motor fault judgment method and device, terminal equipment and computer storage medium |
| CN113255953A (en) * | 2020-11-27 | 2021-08-13 | 中兴通讯股份有限公司 | RRU undervoltage risk prediction method, device, system, equipment and medium |
| CN113887741A (en) * | 2021-11-05 | 2022-01-04 | 深圳市电子商务安全证书管理有限公司 | Data generation method, device, equipment and storage medium based on federal learning |
| CN114092769A (en) * | 2022-01-20 | 2022-02-25 | 国网江西省电力有限公司电力科学研究院 | Transformer substation multi-scene inspection analysis method based on federal learning |
| CN114707606A (en) * | 2022-04-11 | 2022-07-05 | 中国电信股份有限公司 | Data processing method and device based on federal learning, equipment and storage medium |
| CN115130814A (en) * | 2022-05-10 | 2022-09-30 | 中南大学 | Privacy calculation method and system for longitudinal data fusion |
| WO2023103592A1 (en) * | 2021-12-06 | 2023-06-15 | 中兴通讯股份有限公司 | Device risk prediction method, electronic device and computer-readable storage medium |
| CN117094420A (en) * | 2023-10-20 | 2023-11-21 | 浙江大学 | Model training method, device, power prediction method, equipment and medium |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20180314971A1 (en) * | 2017-04-26 | 2018-11-01 | Midea Group Co., Ltd. | Training Machine Learning Models On A Large-Scale Distributed System Using A Job Server |
| CN109189825A (en) * | 2018-08-10 | 2019-01-11 | 深圳前海微众银行股份有限公司 | Lateral data cutting federation learning model building method, server and medium |
| CN109492420A (en) * | 2018-12-28 | 2019-03-19 | 深圳前海微众银行股份有限公司 | Model parameter training method, terminal, system and medium based on federation's study |
-
2019
- 2019-12-26 CN CN201911365903.9A patent/CN111143308B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20180314971A1 (en) * | 2017-04-26 | 2018-11-01 | Midea Group Co., Ltd. | Training Machine Learning Models On A Large-Scale Distributed System Using A Job Server |
| CN109189825A (en) * | 2018-08-10 | 2019-01-11 | 深圳前海微众银行股份有限公司 | Lateral data cutting federation learning model building method, server and medium |
| CN109492420A (en) * | 2018-12-28 | 2019-03-19 | 深圳前海微众银行股份有限公司 | Model parameter training method, terminal, system and medium based on federation's study |
Non-Patent Citations (1)
| Title |
|---|
| 谢丰;卞建玲;王楠;郑倩;: "联邦学习在泛在电力物联网人工智能领域的应用" * |
Cited By (29)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111678696A (en) * | 2020-06-17 | 2020-09-18 | 南昌航空大学 | Intelligent mechanical fault diagnosis method based on federal learning |
| CN111737749A (en) * | 2020-06-28 | 2020-10-02 | 南方电网科学研究院有限责任公司 | Measuring device alarm prediction method and device based on federal learning |
| CN111723942A (en) * | 2020-06-29 | 2020-09-29 | 南方电网科学研究院有限责任公司 | Enterprise power load prediction method, power grid service subsystem and prediction system |
| CN111722043A (en) * | 2020-06-29 | 2020-09-29 | 南方电网科学研究院有限责任公司 | Power equipment fault detection method, device and system |
| CN111723942B (en) * | 2020-06-29 | 2024-02-02 | 南方电网科学研究院有限责任公司 | Enterprise electricity load prediction method, power grid business subsystem and prediction system |
| CN111722043B (en) * | 2020-06-29 | 2021-09-14 | 南方电网科学研究院有限责任公司 | Power equipment fault detection method, device and system |
| CN111796811B (en) * | 2020-07-08 | 2021-07-30 | 上海富数科技有限公司 | Flow control engine system for realizing support of breakpoint debugging in federal learning |
| CN111796811A (en) * | 2020-07-08 | 2020-10-20 | 上海富数科技有限公司 | Flow control engine system for realizing support of breakpoint debugging in federal learning |
| CN111856934B (en) * | 2020-07-16 | 2022-11-15 | 南京大量数控科技有限公司 | Federal learning data processing algorithm between isomorphic intelligent workshops |
| CN111856934A (en) * | 2020-07-16 | 2020-10-30 | 南京大量数控科技有限公司 | Federal learning data processing algorithm between isomorphic intelligent workshops |
| CN112085074A (en) * | 2020-08-25 | 2020-12-15 | 腾讯科技（深圳）有限公司 | Model parameter updating system, method and device |
| CN112085074B (en) * | 2020-08-25 | 2024-05-07 | 腾讯科技（深圳）有限公司 | Model parameter updating system, method and device |
| EP4242939A4 (en) * | 2020-11-27 | 2024-05-15 | Zte Corp | Rru undervoltage risk prediction method, apparatus, and system, device, and medium |
| CN113255953A (en) * | 2020-11-27 | 2021-08-13 | 中兴通讯股份有限公司 | RRU undervoltage risk prediction method, device, system, equipment and medium |
| WO2022111068A1 (en) * | 2020-11-27 | 2022-06-02 | 中兴通讯股份有限公司 | Rru undervoltage risk prediction method, apparatus, and system, device, and medium |
| CN112763845A (en) * | 2020-12-23 | 2021-05-07 | 广东电网有限责任公司梅州供电局 | Edge Internet of things firmware fault detection method and system based on federal learning |
| CN112763845B (en) * | 2020-12-23 | 2022-07-08 | 广东电网有限责任公司梅州供电局 | Edge Internet of things firmware fault detection method and system based on federal learning |
| CN113033605A (en) * | 2021-02-08 | 2021-06-25 | 广东迅科动力科技有限公司 | Motor fault judgment method and device, terminal equipment and computer storage medium |
| CN113887741A (en) * | 2021-11-05 | 2022-01-04 | 深圳市电子商务安全证书管理有限公司 | Data generation method, device, equipment and storage medium based on federal learning |
| CN113887741B (en) * | 2021-11-05 | 2022-09-30 | 深圳市电子商务安全证书管理有限公司 | Data generation method, device, equipment and storage medium based on federal learning |
| WO2023103592A1 (en) * | 2021-12-06 | 2023-06-15 | 中兴通讯股份有限公司 | Device risk prediction method, electronic device and computer-readable storage medium |
| CN114092769A (en) * | 2022-01-20 | 2022-02-25 | 国网江西省电力有限公司电力科学研究院 | Transformer substation multi-scene inspection analysis method based on federal learning |
| CN114092769B (en) * | 2022-01-20 | 2022-06-14 | 国网江西省电力有限公司电力科学研究院 | Transformer substation multi-scene inspection analysis method based on federal learning |
| CN114707606A (en) * | 2022-04-11 | 2022-07-05 | 中国电信股份有限公司 | Data processing method and device based on federal learning, equipment and storage medium |
| CN114707606B (en) * | 2022-04-11 | 2023-12-22 | 中国电信股份有限公司 | Data processing method and device based on federal learning, equipment and storage medium |
| CN115130814B (en) * | 2022-05-10 | 2023-05-02 | 中南大学 | Privacy computing method and system for longitudinal data fusion |
| CN115130814A (en) * | 2022-05-10 | 2022-09-30 | 中南大学 | Privacy calculation method and system for longitudinal data fusion |
| CN117094420A (en) * | 2023-10-20 | 2023-11-21 | 浙江大学 | Model training method, device, power prediction method, equipment and medium |
| CN117094420B (en) * | 2023-10-20 | 2024-02-06 | 浙江大学 | Model training method, device, power prediction method, equipment and medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN111143308B (en) | 2023-04-28 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111143308B (en) | High-low voltage motor data processing method, system and device based on federal learning | |
| US9401839B2 (en) | Generation and control of network events and conversion to SCADA protocol data types | |
| Davis et al. | A cyber-physical modeling and assessment framework for power grid infrastructures | |
| Dong et al. | Numerical simulation for stochastic transient stability assessment | |
| CN109921573A (en) | A kind of system that extensive motor predictive maintenance is realized based on edge calculations gateway | |
| US20190228110A1 (en) | System and method for abstracting characteristics of cyber-physical systems | |
| CN105242128A (en) | Online lightning arrester monitoring system base on fuzzy neural network | |
| CN116599776B (en) | Smart electric meter management method, device, equipment and storage medium based on Internet of things | |
| CN104895547A (en) | Intelligent mobile monitoring and full-automatic maintenance method and device of electric control system of drilling machine | |
| CN107762821A (en) | A kind of air compressor monitoring method and system based on Internet of Things | |
| CN104375493A (en) | Intelligent monitoring system and method for high-voltage motor | |
| US20230118175A1 (en) | Event analysis in an electric power system | |
| CN106973034A (en) | System and method for the data of connection object | |
| CN104238521A (en) | Offshore wind plant remote management system | |
| Daniar et al. | Optimal controlled islanding considering frequency‐arresting and frequency‐stabilising constraints: A graph theory‐assisted approach | |
| US9581995B2 (en) | Systems and methods for monitoring of equipment using a configurable rule base | |
| WO2023240280A1 (en) | Systems and methods for anomaly detection | |
| CN116467592A (en) | Production equipment fault intelligent monitoring method and system based on deep learning | |
| CN110855757A (en) | Soft start equipment management system and method based on cloud service | |
| CN113268844B (en) | Fault equipment acquisition method, device and equipment for power line | |
| CN115437886A (en) | Fault early warning method, device and equipment based on storage and calculation integrated chip and storage | |
| CN108600155A (en) | A kind of convergence-level network security prevents the industrial control system invaded outside | |
| CN111539642B (en) | Object-oriented power plant data acquisition and processing system and method thereof | |
| CN204754902U (en) | Control of rig electrical system intelligent movement and full -automatic device of maintaining | |
| CN114363347A (en) | Self-adaptive industrial equipment data acquisition method and system |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |